Integrating data-driven methods and expert knowledge to develop personas: Balancing automation and multi-disciplinary validation

Abstract

Data-driven personas are increasingly used to inform design decisions. Various methods are published to produce personas based on data collected from projects of different types and scales, each with a specific focus. This study aims to create a set of personas using data collected from a prior randomised controlled trial (RCT), which will be instrumental in designing future eHealth applications to support individuals with cardiovascular disease (CVD). Our method followed five phases for designing personas: (Phase I) expert analysis and variable selection, (Phase II) clustering, (Phase III) expert validation, (Phase IV) persona optimisation, and (Phase V) final check. To ensure that personas accurately reflected the patients, we employed the k-prototype algorithm to cluster mixed data and we focused on validation with colleagues, including medical colleagues, physiotherapists, a psychologist and Human-Computer Interaction (HCI) experts. Seven different personas resulted from the clustering. A validation step involved a multidisciplinary team that assessed the personas' realism, giving an average rating of 8.0 out of 10. Based on their feedback, three of the personas were slightly updated. The final descriptions of all seven personas incorporated the clustered data and the proposed changes after the validation. We concluded that data-driven approaches and expert-based refinement to develop personas is an effective method for understanding the target population. This study highlighted the importance of validation, revealing that creating personas cannot be fully automated, as this may result in losing essential characteristics that only experts can identify. Future research includes demonstrating the practical use of personas.